Orthogonal Routing Protocol

1
Orthogonal Routing Protocol

• Bow-Nan Cheng

2
ORP Overview

• Introduction / Basic Principles
• Design Goals / Key Advantages
• Protocol Specifications
• Analysis and Basic Simulations
• Packetized Simulations and Evaluation
• Conclusion and Future Work

3
Introduction

• Orthogonal Routing Protocol (ORP) is a
  lightweight routing protocol utilizing
  directional communications methods (such as
  directional antennas or free-space-optics LEDs)
  to relax assumptions of location to address
  mapping while providing for efficient medium
  reuse. ORP provides connectivity under extreme
  conditions of high-speed mobility, connectivity
  disruptions and minimal information (ie lack of
  GPS localizations etc.) with relatively high
  spatial reuse.

4
ORP Basic Concepts

• Assuming directional transmissions either through
  FSO transceivers or directional antennas
• Drawing two pairs of perpendicular lines
  intersecting at two different nodes will always
  yield atleast 1 intersection in each direction
  between the perpendicular line pairs (see
  picture)

5
ORP Basic Concepts

• In many cases, nodes will have different sense of
  direction (ie Node D thinks north is to the
  right while Node A thinks north is up)
• Orthogonal intersections will yield a rendezvous
  point/node regardless of individual nodes sense
  of direction
• Supposing Node A wishes to send to Node D, the
  path taken would be through the
  intersection/rendezvous node as show in picture

6
ORP Design Goals

• Provide Connectivity Under Lessened / Relaxed
  Information
• Lack of unified location discovery service
• Lack of universal position and orientation
• Efficient Medium Reuse
• Directionality of sending ensures security and
  less collisions
• Highly Scalable
• Less state information maintained at each node

7
ORP Tradeoffs

• ORP Path not necessarily Shortest Path
• ORPs path selection can be suboptimal
• Need to show that ORP Path deviation, under
  normal-use conditions, are acceptable when
  compared to shortest path
• ORP in its purest form, might result in
  unreachability of some nodes.
• ORPs rendezvous node could be outside of
  topology area
• Need to show that these cases are extreme or
  propose an alternative (future work)

8
ORP Specs Assumptions

• Neighbor Discovery (MAC Property)
• Local Sense of Direction
• Ability to Transmit Directionaly (Antenna
  Property)

9
ORP Specs Theory

• Key Questions
• What are the proactive elements of the routing
  protocol? Why are they necessary and what aspects
  can be tweaked based on usage scenario?
• What are the reactive elements of ORP and why
  were some design decisions chosen over others?
• How do we address issues with non-ideal
  situations where packets stray from being
  forwarded in the orthogonal direction?
• How does ORP recover from route errors and
  prevent potential routing loops?
• What are some ways ORP deals with sparse and
  highly mobile environments?

10
ORP Specs Proactive Elements

• Goal Establish Rendezvous-node-to-destination
  routes
• Tweak Specs Announcement interval, Route timeout

11
ORP Specs Proactive Elements

• ORP Announcements
• Using local sense of direction, each node sends
  an ORP announcement packet in orthogonal
  directions
• ORP Forwarding Table Build
• Upon receipt of ORP Broadcast PKT, build
  forwarding table (NB ID neighbor ID, Dir
  Direction, Dest ID Source of ORP Broadcast)
• If TTL on ORP Broadcast Packet not expired, send
  packet along path in same direction (send
  transceiver is 180 degrees from receive
  transceiver)

12
ORP Specs Reactive Elements

• ORP Route Request and Route Reply
• When a node receives an ORP RREQ, check to see if
  we have the destination ID in our forwarding
  table
• If destination ID not in our forwarding table
• Add 180 degrees to the orientation of the
  transceiver that received it and send out of
  nearest transceiver
• Else
• Send ORP Probe ACK packet back through
  transceiver we received from

13
ORP Specs Reactive Elements

• Goal Establish the Source-to-rendezvous-node
  route
• Source Based routing vs. Next Hop routing
• Advantages of Next Hop
• Less header information
• Easy to update routes
• Forwarding table search easy
• Advantages of Source-Based
• Can potentially encode trajectory into packet
• New paths can be encoded into packet header
  quickly

14
ORP Specs Forwarding Tables

• Note only 1-hop tables maintained. Because of
  nature of ORP broadcast, forwarding tables might
  not include all immediate 1 hop neighbors

15
ORP Specs Data Forwarding

• Data Packet Forwarding
• - When a node receives an Data Packet, check to
  see if we have the destination ID in our
  forwarding table
• - If destination ID not in our forwarding table
• - Add 180 degrees to the orientation of the
  transceiver that received it and send out of
  nearest transceiver
• Else
• Send it in the direction of destination

16
ORP Specs Packet Deviation Correction

• Problem real-world cases usually dont have
  nodes aligned in grid-like fashion. Deviation in
  forwarding along a line can severely mess up ORP
  forwarding
• In example shown, source node S, though intending
  to send in orthogonal directions, ends up sending
  in directions far from orthogonal.

17
ORP Specs Packet Deviation Correction

• Three Step, Three State Method
• Each RREQ and Announcement packet maintains 2
  additional states in packet headers
• Deviation
• Deviation State
• Initial deviation is recorded in deviation
  variable in packet header and deviation state set
  to 1
• After 2nd node receives packet and sees deviation
  state set to 1, forward packets in opposite
  direction of receipt 2deviation and set
  deviation state to 2
• After 3rd node receives packet, forward packets
  in opposite direction of receipt deviation and
  set deviation state back to 1

18
ORP Specs Packet Deviation Correction Example
19
ORP Specs Loops and Mobility

• ORP assumes no loops because all packets are
  being forwarded in opposite directions ensuring
  forward forwarding
• In highly mobile environments, route entry
  timeout is reduced and announcement interval
  frequency increased. Because of the
  directionality of the sending signal, medium
  reuse is ok

20
ORP Analysis and Basic Simulations

• Key Questions
• Are there certain conditions in which ORP cannot
  successfully deliver a packet? What is the upper
  bound on the source-destination reachability in
  ORP?
• How much state information does each node on
  average need to maintain and how does this
  compare with other protocols in terms of
  scalability?
• How inefficient is ORP path selection compared to
  shortest path. Are these results an acceptable
  sacrifice for connectivity?

21
ORP Reach Probability Upper Bound

• Methodology Orthogonal lines drawn from source
  and destination. Intersection points measured and
  if all intersection points outside of topology
  area, then destination is unreachable given
  orientation and position of source and destination

22
ORP Total State Information

• To measure scalability, it is important to see
  how much state information needs to be maintained
  in the network. ORP was found to need to maintain
  order N3/2 states even for varying topologies

23
ORP Shortest Path Inefficiency

• The ORP Path selected vs. Shortest path for a
  number of topologies was measured. On average, in
  symmetrical topologies like a circle or square,
  ORP Path was comparable to Shortest path.

24
ORP Packetized Simulations

• Network Simulator used for simulations
• Key Questions
• What amount of state is maintained in ORP under
  random topologies and real-world conditions and
  how does it compare with other protocols?
• What is the level of connectivity in random
  networks using ORP?
• What happens to ORP when mobility enters the
  equation? Are packets still able to be
  successfully delivered?
• How does ORP's sending of control packets compare
  with DSDV,AODV, and other routing protocols?

25
ORP NS2 Total State vs. Number of Nodes

• Under grid and random topologies, the total state
  maintained in the network vs. the number of nodes
  was measured. As expected, the fitted curve was
  order N3/2

26
ORP NS2 Frequency of State Information
27
ORP NS2 Spread of State Information in Topology

• In the grid topology, the edge nodes seem to
  maintain more state. This is only somewhat the
  case in random topologies